Research On Intrusion Detection Technology Based On Convolutional Neural Network

Intrusion detection technology is the key technology in network security,with the diversification of means of network attacks,the traditional intrusion detection technology has gradually revealed some problems,such as poor detection performance and low adaptability.In view of the existing problems in intrusion detection,this paper constructs a model of Intrusion Detection Based on convolution neural network,and the convolution neural network is further studied and improved.The main works of this thesis are as following:(1)This paper optimizes the initial weights of convolutional neural network by genetic algorithm on the problems of slow training speed and difficult convergence in the training of convolutional neural network.The experimental results show that the convergence speed of the convolution neural network optimized by genetic algorithm is faster and the feature extraction ability is strengthened.(2)On the problem of low detection performance in current intrusion detection,this paper applies deep convolution neural network with excellent feature extraction ability to intrusion detection,and improves the convolution neural network model by adjusting the network structure,increasing the number of convolution kernel,simplifying the full connection layer,adjusting the learning rate and other methods.The experimental results show that the improved convolution neural network has good ability to identify the characteristics of abnormal data.(3)This paper designs an intrusion detection model based on convolutional neural network and uses the KDDCUP 99 data set to test the model.The intrusion detection model has good performance on three indexes,the accuracy rate of intrusion detection,false alarm rate and false negative rate,it also has a high recognition rate for new attacks.The experimental results show that the intrusion detection model based on convolutional neural network can detect various kinds of abnormal data and attack types effectively,and also has the ability to detect new attack data.